Genera models

Of the 182 identified genera of importance for deer and elk forage or sage-grouse, we had sufficient data (e.g. at least 3 ecognition polygons) to train models for 153 of the genera

In the vast majority of cases, Random Forest models appeared to train the best predictive models based on AUC values.

BestModel Number of Genera
RF 89
XGBOOST 26
GBM 12
GLMNET 12
GLM 7
GLMNET_class 6
RF,XGBOOST 1

We were able to attain AUC values of >= 0.8 for 76% of the trained genera models.

AUC values plotted according to sample size

Data summary of trained models.

Species models

Of the 593 identified species of importance for deer and elk forage or sage-grouse, we had sufficient data (e.g. at least 3 ecognition polygons) to train models for 240 of the species.

In the vast majority of cases, Random Forest models appeared to train the best predictive models based on AUC values.

BestModel Number of Species
RF 111
XGBOOST 51
GLMNET 32
GBM 19
GLMNET_class 17
GLM 8
GLM,RF 1
RF,XGBOOST 1

We were able to attain AUC values of >= 0.8 for 85% of the trained species models.

AUC values plotted according to sample size

Data summary of trained models.

Subspecies models

Of the 216 identified subspecies of importance for deer and elk forage or sage-grouse, we had sufficient data (e.g. at least 3 ecognition polygons) to train models for 9 of the subspecies.

BestModel Number of Subspecies
RF 4
GBM 2
XGBOOST 2
GLMNET 1

We were able to attain AUC values of >= 0.8 for 78% of the trained subspecies models.

AUC values plotted according to sample size

Data summary of trained models.